Rumours out that Kepler GK104 cards will be $299...

[Jokester]

This could be the 6870 all over again, people think it replaces the 5870 but its really just a mid range side-grade. Best to see the full range before picking the best price point for yourself.

This is my thoughts, I don't believe Nvidia will release their top card at £240. For me this points faster cards following quickly on rather than the "anything upto Q4" rumours that are going round.

I also believe that this points to the hot shader clocks being done away with so performance estimates will be wildly optimistic.

 

[xsistor]

There are some rumours that there will be a GTX 660 and a GT 680 coming out.

GT 680!

GTX 680 will come later

 

[TheMorningStar]

2gb of vram and at that price and gtx580 like performance is a big win for nvidea and us. Here's hoping.

 

[ICDP]

2gb of vram and at that price and gtx580 like performance is a big win for nvidea and us. Here's hoping.

IF it ends up with GTX 580 like performance for $299 I could see AMD responding by reducing the HD 7950 1.5GB versions to a similar, or slightly higher price.

If this is the price/performance we are to excpect then I would imagine a GT 680 card would be similar in speed to an overclocked HD 7970. AMD would have no choice but to reduce prices to match price/performance to Nvidia's offerings.

 

[ALXAndy]

Wow a launch card that's comparable to the 580 ! how exciting. Not.

Might bring the prices of the 7970 down a little, but unless it beats it then AMD can just continue to charge what they like for it, given it sounds like the 7970 is still going to be the world's fastest single GPU video card and thus carries the elite price.

 

[dave_beast]

IF it ends up with GTX 580 like performance for $299 .......

I would imagine a GT 680 card would be similar in speed to an overclocked HD 7970.

A standard clocked 7970 beats a GTX580, and overclocked 7970 destroys it, so you're argument is faulty.

AMD have got it right imo, they released their fastest single GPU first, then the lesser models later on. For those wanting the ultimate, AMD is the only route atm. If NVidia release a mid~high card before ther top one, it's pretty much forcing folks to go AMD.

 

[ALXAndy]

AMD have got it right imo, they released their fastest single GPU first, then the lesser models later on.

Yup, that.

And they have been very, very clever with it too.

Where are the cheap cards that can be ran in pairs and offer the same performance as the 7970 for a lot less cash?

There is no temptation this time. They're enjoying their place atop the podium for a while.

Ed. And all signs point to the GPU in the 7970 having more left in the tank, too. Something was locked on it, sure I read that somewhere.

 

[xsistor]

Wow a launch card that's comparable to the 580 ! how exciting. Not.

Might bring the prices of the 7970 down a little, but unless it beats it then AMD can just continue to charge what they like for it, given it sounds like the 7970 is still going to be the world's fastest single GPU video card and thus carries the elite price.

Always the cynic...

 

[ALXAndy]

Always the cynic...

Oh come off it man. It's hardly going to be revolutionary is it?

I just bought a 7970. It would take something faster to turn my eye and possibly force me into a Uturn.

So you'll have to excuse me for not being very excited that the next gen of Nvidia cards won't be competing at the high end for a while.

I'd be mad to be excited about something slower than what I already have, and, by the time of the release of Kepler will have had for a while.

 

[Havana_UK]

A standard clocked 7970 beats a GTX580, and overclocked 7970 destroys it, so you're argument is faulty.

AMD have got it right imo, they released their fastest single GPU first, then the lesser models later on. For those wanting the ultimate, AMD is the only route atm. If NVidia release a mid~high card before ther top one, it's pretty much forcing folks to go AMD.

I both agree and disagree.

AMD have got it right releasing their top-end card first. They've claimed the single GPU performance crown, they've given themselves the opportunity to charge a premium price for their product, and those wanting the ultimate (right now) have to choose AMD.

However, nVidia releasing a mid range part first was probably (certainly?) not in the original plan, and to me it seems like a reaction to how well Tahiti performs, and an attempt to retain market share in the mid-high range (aka against 7950, NOT 7970) where - let's face it - the volumes are substantially higher. The high end cards (single and multi-GPU solutions) are only an exercise in "willy waving", and while nVidia are doubtless gutted to be behind AMD in performance it has little real impact on the overall company performance.

If the GK104 based card performs as well as, or slightly better than, the GTX 580 that will be a significant advantage for nVidia. It's generally believed that nVidia have two higher-end parts in the pipeline, GK110 and GK112, and if GK104 performs as well as hoped this leaves us to expect trully staggering things from the high-end parts that will dislodge 7970's performance crown with a huge thump.

7970 does beat GTX 580 sometimes significantly, sometimes less so, but it really only begins to shine out when clocked way above the standard (which is does phenomenally well, I'm sure all will agree). If GK104 is equivalent or slightly better than GTX 580 then the gap to 7970 isn't huge, and who's to say how well GK104 will clock. Maybe that will be a monster as well?

GK104 is definitely intended to compete with 7950, but if it ends up getting close to 7970 performance then AMD will really struggle against GK110/112.

 

[xsistor]

Oh come off it man. It's hardly going to be revolutionary is it?

I just bought a 7970. It would take something faster to turn my eye and possibly force me into a Uturn.

So you'll have to excuse me for not being very excited that the next gen of Nvidia cards won't be competing at the high end for a while.

I'd be mad to be excited about something slower than what I already have, and, by the time of the release of Kepler will have had for a while.

Well at least you didn't tell us why our opinion was bull and then write 5 pages on your own opinion... To your credit.


Anyway, the GTX 660 is at least as fast as a 580. But there are rumors it will match a 7970. And at $299, that is something to be excited about indeed. Because it shows that the higher end cards will be much more powerful... And not so wildly priced. So yeah, it's exciting news to me, if true.

 

[Biffa]

Is all this because effectively Nvidia are doing a "tick" and AMD are doing a "tock"?

 

[ALXAndy]

7970 does beat GTX 580 sometimes significantly, sometimes less so, but it really only begins to shine out when clocked way above the standard (which is does phenomenally well, I'm sure all will agree). I

That will change. It's clear from looking at performance figures in older games that Tahiti was rushed out of the door. Newer games beat the 580 quite easily, and that's because they just haven't had the time to improve the drivers on older (and what could be seen as) less significant games.

Right now there is the big BF3 frenzy. That is the talking point, that is where you want your card to be better. That would be why people have bought a lot of cards recently (just looking around suggests that my reasoning has weight).

And quelle surprise the latest drivers improve BF3 performance.

 

[xsistor]

I both agree and disagree.

AMD have got it right releasing their top-end card first. They've claimed the single GPU performance crown, they've given themselves the opportunity to charge a premium price for their product, and those wanting the ultimate (right now) have to choose AMD.

However, nVidia releasing a mid range part first was probably (certainly?) not in the original plan, and to me it seems like a reaction to how well Tahiti performs, and an attempt to retain market share in the mid-high range (aka against 7950, NOT 7970) where - let's face it - the volumes are substantially higher. The high end cards (single and multi-GPU solutions) are only an exercise in "willy waving", and while nVidia are doubtless gutted to be behind AMD in performance it has little real impact on the overall company performance.

If the GK104 based card performs as well as, or slightly better than, the GTX 580 that will be a significant advantage for nVidia. It's generally believed that nVidia have two higher-end parts in the pipeline, GK110 and GK112, and if GK104 performs as well as hoped this leaves us to expect trully staggering things from the high-end parts that will dislodge 7970's performance crown with a huge thump.

7970 does beat GTX 580 sometimes significantly, sometimes less so, but it really only begins to shine out when clocked way above the standard (which is does phenomenally well, I'm sure all will agree). If GK104 is equivalent or slightly better than GTX 580 then the gap to 7970 isn't huge, and who's to say how well GK104 will clock. Maybe that will be a monster as well?

GK104 is definitely intended to compete with 7950, but if it ends up getting close to 7970 performance then AMD will really struggle against GK110/112.

Actually, Havana, all rumors from the beginning pointed to Kepler coming out feet first. The lower end cards were set to be released first and then followed by the higher end. One of the slides showed the GTX 680 in the latter half of this year and then a super high-end GK112/512-bit card in Jan 2013.

And according to the present rumors, the reason NVIDIA has switched this to go mid-range first, followed by high end is because they feel their midrange will be competitive against the 7970 and they don't want AMD to rule for too long.

 

[ALXAndy]

Well at least you didn't tell us why our opinion was bull and then write 5 pages on your own opinion... To your credit.


Anyway, the GTX 660 is at least as fast as a 580. But there are rumors it will match a 7970. And at $299, that is something to be excited about indeed. Because it shows that the higher end cards will be much more powerful... And not so wildly priced. So yeah, it's exciting news to me, if true.

People are being deluded again.

If it matches a 7970 (and I would be a lot more excited if it did !) then it won't cost $299. It's never happened before and it won't happen any time soon, especially out of Nvidia.

*IF* (and the magic 8 ball says "All signs point to yes") it is the replacement part for the 580 it will cost around the same. So around £300-£350.

Plus Nvidia have Physx CUDA and all of their other pointless Swiss Army things aboard so they will charge a little extra for those.

But any one who was thinking Nvidia are going to release a card that sits in the top end scales for $299 needs to go and take a cold shower and sit in the corner for a while.

 

[xsistor]

PhysX and CUDA are far from pointless.

 

[drunkenmaster]

Actually, Havana, all rumors from the beginning pointed to Kepler coming out feet first. The lower end cards were set to be released first and then followed by the higher end. One of the slides showed the GTX 680 in the latter half of this year and then a super high-end GK112/512-bit card in Jan 2013.

And according to the present rumors, the reason NVIDIA has switched this to go mid-range first, followed by high end is because they feel their midrange will be competitive against the 7970 and they don't want AMD to rule for too long.

That's not exactly accurate, all the rumours SINCE the rumours the top end part hasn't taped out yet say essentially that now its been delayed....... what is coming out first.

The name pretty much suggests it all as well in the rumours.

If the Gk110 or GK112 rumour is true. Every other generation = lower the number higher the part, IE GK100 would have been the likely "true" high end part" and GK110 would have been the likely "refresh" part. 480gtx = gf100, 580gtx = gf110.

IE the fact that the rumours are all talking about a GK110/112 for the high end part in and of itself indicates the GK100 part got canceled/delayed or just didn't work at the time.

 

[MichaelM]

First page I will go to on reviews of the GK104 will be power consumption, that will set the tone for its success and the rest of the lineup.

 

[ALXAndy]

PhysX and CUDA are far from pointless.

They're hardly mainstream are they?

If they had a major worthwhile point people wouldn't buy any AMD cards.

Yeah, there are some pretty cool Physx titles out there. However, Physx does absolutely nothing Havok can't. And Havok can do it with an API using your CPU.

The creators of Havok realised this, and instead of wasting time like Ageia they decided to market it as an API instead.

Nvidia insisted on bolting it to their cards, heavily decreasing their target audience. Which is why we have few decent Physx titles because you need to be in complete cahoots with Nvidia in order to do so, which limits your target audience making it a brave thing to do.

CUDA? another server workstation technology that is of very little use to gamers. Good for folding, mind, but again you are reaching out to a small target audience.

 

[Havana_UK]

Is all this because effectively Nvidia are doing a "tick" and AMD are doing a "tock"?

Where'd you get that from?

4xx was a "tock" and 5xx was a "tick", assuming you mean that the tocks are the architectural changes and the ticks are refreshes (ala Intel).

Kepler is clearly a "tock". 7970 is also a "tock".